Part of the Dashlibs suite — Databricks libraries built for business users.
ML lifecycle management: preprocessing, drift monitoring, evaluation (SHAP + model cards), governance artifacts, and champion/challenger promotion — driven from one notebook UI, backed by Unity Catalog and MLflow.
%pip install dash-mlopsimport dashml
dashml.launch() # Opens interactive UI in your Databricks notebook| Area | Entry points |
|---|---|
| Preprocessing | clean_dataframe(), dashml.transforms (outlier removal, binning, lag features, ...) |
| Drift monitoring | ModelMonitor (PSI + chi-squared, optional auto-retrain trigger) |
| Evaluation | explain_features() (SHAP), build_model_card(), check_thresholds() |
| Governance | build_governance_artifacts() (signature, features, fairness, approval record) |
| Registry | RunTracker, register_model(), promote_challenger() (UC @champion alias) |
| Experimentation | dashml.experiment — compare/promote MLflow runs |
| Serving | dashml.serving.sync_serving_endpoint() |
Everything beyond ModelMonitor and the notebook UI is also directly
importable for use in a training script — launch() is the guided path,
not the only path.
| Library | Purpose |
|---|---|
| dash-dq | Data Quality |
| dash-synthetic | Synthetic Data Generation |
| dash-ml | ML Lifecycle Management |
| dash-ingest | Data Ingestion |
| dash-gov | Data Governance |
| dash-ontology | Ontology & Lineage for AI |
Apache 2.0